What you’ll leave with
A three-scene storyboard, a precise animation brief, and a repeatable way to check an exported explainer.
Before you begin. A Claude Team or Enterprise account with Motion enabled, plus a small report or dataset you are allowed to use. The walkthrough is an original exercise, not a claim that we rendered the video in Claude.
What the documentation establishes
Anthropic describes Motion as a beta for Team and Enterprise. It creates code-driven animations rather than photorealistic generated footage. Source ↗
The documented entry points are Output → Motion in chat or a Motion template in Artifacts; the documented download format is MP4. A /motion command is not required by this guide. Source ↗
Motion is enabled by default for Team and disabled by default for Enterprise, where an owner controls access in Organization settings → Artifacts. Source ↗
The walkthrough below is an original worked exercise. We have not independently tested this product’s performance.
A useful explainer changes what the audience understands. Moving a paragraph across the screen is rarely enough. Start with one question: after watching, what should somebody be able to explain or do? For this exercise, the audience is a product team deciding whether to investigate an onboarding problem.
We will use deliberately invented numbers: 1,000 people start onboarding, 640 reach profile setup, and 410 finish. Keep the word ‘example’ on screen. The goal is to make the biggest loss visible without implying that the animation has diagnosed its cause. A chart can show where people leave; it cannot tell you why.
1. Choose the right kind of moving image
Use this exercise when the content is mostly labels, diagrams, charts, or product steps. If the brief depends on a believable actor speaking to camera, capture real footage or evaluate a tool made for that job. If the audience only needs a number, a static chart may communicate it faster than a video.
Write the target surface before generating: a widescreen meeting display, a vertical social clip, or a small help panel. Make a separate composition for each. Resizing a widescreen chart into a tall frame usually produces tiny labels and wasted space. For the first attempt, choose one surface and one audience.
2. Make the story work before adding motion
| Time | What appears | What the viewer learns |
|---|---|---|
| 0–6 seconds | ‘Example onboarding funnel’ and the three stage names | What the numbers describe |
| 6–22 seconds | Reveal 1,000 → 640 → 410, keeping the scale consistent | The largest absolute loss is before profile setup |
| 22–30 seconds | ‘Next: inspect the first transition’ with one research question | A specific investigation, not an unsupported causal claim |
Read the storyboard aloud without any animation. If you cannot explain it in the allotted time, remove a point. Treat holds as part of the timing: the final recommendation needs time on screen. Do not replace a readable hold with a dramatic transition simply because the tool can generate it.
3. Give Claude a constrained production brief
Open an available Motion entry point and attach the source file. Put the authoritative figures in the prompt too, so discrepancies are easy to spot. Ask for the storyboard first; correcting three scene descriptions is cheaper than repeatedly revising a finished animation.
Create a 30-second, 16:9 onboarding explainer for a product-team meeting.
This is an illustrative dataset, not company performance:
start=1000; profile=640; complete=410.
First return a three-scene storyboard with durations and exact on-screen text.
Keep the label "Illustrative data" visible. Use one consistent chart scale.
Do not infer causes, add numbers, logos, voice claims, or customer quotes.
The final action is: inspect the transition from start to profile setup.
Use readable labels, restrained transitions, and a final hold.
After I approve the storyboard, build the animation from that plan.4. Review one kind of error at a time
Run a factual pass first. Compare every numeral, unit, and label with the supplied data. In this example, a label saying ‘59% drop at profile’ would be misleading: the complete funnel loses 59%, while the first transition loses 36%. Ask for the calculation behind any annotation.
Then run a readability pass at the size your audience will actually see. Pause each scene. If you need to zoom to read it, shorten the label or simplify the chart. Finally inspect timing. Give one precise edit, such as keeping the chart stationary for three more seconds, rather than asking to ‘make it more polished.’
Keep all figures, chart scales, colors, and scene order unchanged. In scene 2, hold the final three bars for four seconds before the transition. Return the revised timing table and identify any text you changed.5. Judge the downloaded file, not just the preview
Use the documented MP4 export and play the saved file from beginning to end. Check the first frame, the last hold, cropped edges, font rendering, and whether the explanation still makes sense without sound. Save the storyboard, source data, and final brief alongside the export so the next revision has a reliable starting point.
Stop iterating when the video communicates the single intended decision and the acceptance checks pass. If each edit breaks a previously correct number, return to a simpler storyboard. A short, accurate diagram is more useful than an elaborate clip nobody can verify.
| Problem | What to try |
|---|---|
| Motion is missing | Check plan and the organization's Artifacts setting before rewriting prompts. |
| The result invents a causal explanation | Remove interpretation from the brief; ask for observable values and explicitly labelled hypotheses. |
| An edit changes unrelated scenes | Restate the protected figures and specify one scene and one property to change. |
| The exported video is hard to read | Test at delivery size; simplify the layout rather than exporting the same crowded composition again. |
Sources & next steps
- Claude Motion setup ↗
Capability, entry points, and export documentation; checked October 10.
- Artifacts administration ↗
Workspace access and plan-specific defaults.
Guide history
Revisions preserve the original instructions. A source recheck alone does not create a new revision.
Revision 3 · Added a complete source CSV, denominators and answer checks, documented Motion entry/export paths, prerequisite diagnostics and a revision prompt that preserves the total duration.
Revision 2 · Added usage and sharing facts, a complete release handoff, and a copyable revision note.
Revision 1 · Initial documented guide
Continue learning
Techniques · 11 min read
Skills, subagents, loops, graphs: choose the mechanism your task needs →
Choose skills, workers, retries and explicit state using a worked brief-writing task, then test whether your workflow notices a missing check.
Updated 2026-10-16
Everyday assistants · 7 min read
Meta Muse: give a personal agent a job you can verify →
Check the October 12 Muse model availability, then evaluate a personal-agent task with source evidence and a deliberate memory correction.
Updated 2026-10-13